AI in Finance Summit Fall schedule

Nonlinear AI/ML representations, such as Random Fourier Features, can reshape financial data in powerful ways, but added complexity does not automatically improve forecasting performance.
What matters is not simply how many features or parameters a model has, but how much stable forecasting flexibility the data used can support, after accounting for sample size, feature geometry, and regularization.
In noisy, low-signal financial settings, disciplined benchmarks—regularized linear models and exact kernel methods—remain critical for determining whether model complexity is adding genuine value.

•    Bridging the gap between data teams and business leadership
•    Enabling scalable AI through cross-functional collaboration
•    Using emotional intelligence to lead change and foster adoption
•    Positioning soft skills as a strategic advantage in AI-driven initiatives

Most organizations expect to move the majority of their AI experiments into production within months — but only a quarter have actually gotten there. Financial services isn't immune to that gap; if anything, the institutions in this room are investing hardest into closing it. In this keynote, Lance Senoyuit, Sr. Principal, Financial Services Industry, Executive Advisor, SAP America’s breaks down why the gap between AI ambition and AI in production persists, where the value quietly leaks out along the way, and what it actually takes to turn deployed AI into AI that changes the business.

Key Talking Points:
•    The gap nobody's closing fast enough — more than half of organizations expect to move most of their AI experiments into production within months. Only a quarter have actually gotten there. That gap is where most institutions are sitting right now.
•    Where the value actually leaks out — it's rarely the model. Fragmented data, systems that don't talk to each other, and agents bolted onto old workflows instead of built into them quietly cap what "deployed" AI can deliver.
•    What closing the gap looks like — how leading institutions are turning proprietary data into a strategic asset and building the platform foundation that lets agents operate on real business context, not isolated pilots.

•    Where is AI actually making money, not just generating headlines?
•    How much of your AI spend is real ROI versus expensive experimentation?
•    What is the most overhyped AI use case in finance right now?
•    Why do so many AI pilots look promising but fail to deliver real value at scale?

Panelists: 
Biswa Sengupta, Chief AI Technologist: LLM Suite (CDAO), JPMORGAN CHASE & CO 
Sarah Zeis, Manager, Data Intelligence & Advanced Analytics, PORSCHE FINANCIAL SERVICES 
Lovedeep Saini, Chief Analytics Officer, CONNER STRONG & BUCKELEW 
Sheedsa Ali, Head of Quantitative Equities, METLIFE INVESTMENT MANAGEMENT 
Robin Braun, VP Al Business Development, Hybrid Cloud, HPE

 

•    From models to machines that act; finance is moving from prediction to execution 
•    Autonomy is compounding; what started as automation is becoming decision making 
•    The next edge is not better algorithms; it is systems that think, learn, and act 
•    The future is not AI assisted finance; it is AI driven finance 

•    Is generative AI already becoming commoditized in finance?
•    If everyone has copilots, where is the real competitive edge?
•    What comes after generation; are you ready for AI that actually makes decisions?
•    Who will lead the next wave; incumbents reinventing themselves or AI native players

Panelists:
Vishal Sharma, VP, Software Engineering, BROADRIDGE FINANCIAL
Dhagash Mehta, Head of Applied Machine Learning Research for Investment Management, BLACKROCK MANAGEMENT 
Todd Lyon, VP Engineering Manager, TABBANK 
 

• Why agents fail in front of customers. Most customer-facing AI breaks down not because of the model, but because of the data and context behind it: stale inputs, inconsistent definitions, and answers that can't be traced back to a source.
• Building traceability into every output. How to ground agent responses in governed, versioned data so every answer can be reproduced, audited, and explained, which is critical in regulated financial environments.
• A practical path from pilot to production. The checks, guardrails, and evaluation practices that give compliance, risk, and business teams the confidence to sign off on putting an agent in front of real customers.

•    Memory turns AI from reactive to strategic
•    Reasoning is the edge—prediction alone is not enough
•    Smarter agents make higher-stakes financial decisions
•    The real challenge: trusted memory, controlled reasoning
 

High impact sessions where leading companies showcase real AI use cases in finance, demonstrating how their solutions drive tangible results and business value in practice

1.    Our research says 47% of people in this industry use AI tools their employer has not approved — and 78% of those employers have a policy. So the policy is not the problem. What is?
2.    Is anybody here training a foundation model? And when we say build versus buy — what are we actually still building?
3.    A regulator asks why your system produced a particular answer eighteen months ago. What do you need to show them, and could you show it today?

•    How AI adoption in financial services evolved from experimentation in 2024 to enterprise-wide implementation in 2026
•    Key lessons learned from scaling Generative AI initiatives across compliance, risk, operations, and customer experience
•    What financial institutions must prioritize next to move from AI efficiency gains to long-term competitive advantage

 

High impact sessions where leading companies showcase real AI use cases in finance, demonstrating how their solutions drive tangible results and business value in practice

Latency and errors directly impact critical customer flows, and AI Agents are making those flows more complex
Honeycomb lets you measure the impact of changes and problems in production at an aggregate and detail level using application traces and logs
As a result, you know where agents and AI-deployed code improved, degraded, or broke functionality fast, often before your users notice. 

•    Are traditional stress tests already obsolete in a world of real time AI driven risk?
•    If AI can simulate infinite scenarios, are we finally solving risk or just creating new blind spots?
•    How much real time risk visibility do institutions actually have versus what they claim?
•    When AI reacts faster than humans, who is really in control of risk decisions?

Panelists:
Cristian Homescu, Director, Portfolio Analytics, Chief Investment Office, Global Wealth and Investment Management (GWIM), BANK OF AMERICA MERRILL LYNCH
Andrew McElduff, Vice President, Membership Application Programs & Risk Monitoring Program Operations, FINRA
Akhil Khunger, VP, Quantitative Analytics, BARCLAYS

Modrator: Anant Natekar, Senior Director, Software Engineering, NORTHWESTERN MUTUAL 

•    Is your AI governance truly scaling, or is it already slowing down innovation?
•    How much of your governance is automated versus still relying on manual oversight?
•    Are you building governance for today’s models or for autonomous systems you cannot fully control yet?
•    If governance fails at scale, is it a technology gap or a leadership blind spot?

Panelists: 
Chandni Bhatiam, Vice President, Lead Quantitative Development, J.P. MORGAN
Krishna Chaitanya Yarlagadda, Director- Data Science & AI at MERCURYFINANCIAL
Urie Tucker, Senior Director, Model Risk Oversight, FINRA
Elumalai Rajendran, Vice President | Principal Engineer, US BANK

Moderator: Wanyao Zhang, Product Manager – Credit/Loan, SEAMONEY 

• Building trust and transparency in agentic AI decision-making within financial services

   

• Balancing explainability, governance, and regulatory expectations for autonomous AI systems

   

• Practical approaches to making agentic AI outputs understandable and actionable for stakeholders

 

•    Scaling AI systems reliably in production 
•    Monitoring and troubleshooting multi-agent systems 
•    Best practices and open challenges for production 

 

•    What does agentic AI actually look like in production, and how far are you from it?
•    Where is agentic AI truly outperforming traditional analytics, and where is it still falling short?
•    What broke first when you tried to deploy agentic systems; your tech, your data, or your culture?
•    How much autonomy are you really willing to give AI when real money and risk are on the line?
•    What will it take for you to trust AI to make decisions, not just recommendations?

Panelists: 
Jaydip Mukhopadhyay, Vice President, Data Science and Model Risk, AMERICAN EXPRESS  
Dishti Dave, Site Reliability Engineer - AVP, BARCLAYS  
Aishwarya Kothapally, First Vice President | Financial Crimes Compliance, BHI 
Jennifer Ostyn, Senior Vice President, Revenue, KNIME

Moderator: Cecilia Dones, Adjunct Professor, COLUMBIA BUSINESS SCHOOL

•    How AI is changing traditional FP&A: Its impact on budgeting, forecasting, scenario planning, and reporting, and where human judgment remains essential. 
•    Addressing resistance to adoption: Understanding concerns about job security and reliability, and helping finance teams build confidence in using AI. 
•    Protecting sensitive financial information: Setting clear boundaries around using forecasts, workforce plans, and material nonpublic information with AI tools 

 

•    AI first architecture is not an upgrade, it is a complete rebuild from the ground up
•    Legacy culture slows AI, AI native culture scales it
•    Business models are shifting from products to intelligent, data driven services
•    The gap will not be incremental, it will be exponential between AI native and legacy institutions

 

•    Autonomous agents need guardrails—or they create new risks  
•    Governance must move at the speed of AI 
•    Trust in finance now depends on controlling algorithms  
•    No oversight, no scale: governance is the real enabler